Industries / Industrial equipment and services / HVAC equipment
Controls
Controls are the thermostats, controllers, sensors and built-in software that tell heating and cooling equipment when to run and how hard.
7
Milestones achieved, from 1883 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1883
An electric room thermostat
On 24 July 1883 the US Patent Office granted Warren S. Johnson of Whitewater, Wisconsin, patent 281,884 for an "electric tele-thermoscope". A coil of two metals bent as a room warmed or cooled, closed an electric circuit through cups of mercury, and moved a pointer and rang a bell at a distant point, so an attendant could see from one place whether each room was too warm or too cold. Johnson Controls calls it "a building temperature-control device". In 1885 Johnson and the Milwaukee financier William Plankinton formed the Johnson Electric Service Company, which became Johnson Controls in 1974.
Credit
Warren S. Johnson; William Plankinton
1953
A thermostat anyone can set
Minneapolis-Honeywell put the T-86 "Round" thermostat on sale in 1953. The engineer Carl Kronmiller and the industrial designer Henry Dreyfuss had designed it in 1941, but the project was put on hold during the Second World War. The Minnesota Historical Society says it was easy to make, easy to use and easy to identify, and, citing a 1995 company history, that about one million sold each year. A day-night version with a wind-up timer that lowered the temperature at night followed in 1960, and a model that ran both heating and cooling in 1966.
Credit
Carl Kronmiller and Henry Dreyfuss, for Minneapolis-Honeywell
1995
A shared language for buildings
Building owners needed control systems from different makers to work as one. Makers and engineers first met as a committee at the annual meeting of the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) in June 1987, and in 1995 ASHRAE published Standard 135, BACnet, a communication protocol any maker can build into chillers, air handlers and controllers. It became an international standard, ISO 16484-5, in 2003. BACnet's official site says it "has become the dominant building automation integration protocol globally".
Credit
ASHRAE's BACnet committee; the site names Mike Newman as "the founding father of the BACnet protocol"
2009
Programmable thermostats lose the label
From 1995 the US Environmental Protection Agency's ENERGY STAR program promoted programmable thermostats, which change the setting on a timetable, as a way to cut heating and cooling bills by 10 to 30%. EPA then reviewed five studies that found no statistically significant savings in homes that had them. Many people found them hard to program, and many already turned the dial down by hand. EPA ended its programmable thermostat specification on 31 December 2009, citing "questions concerning the net energy savings". An evaluation of a utility program that gave small businesses free programmable thermostats, based on 194 telephone surveys, found that about two-thirds were already setting back their old thermostats by hand, so most of the expected savings never came.
LessonA control saves energy only if people use it as intended, so its savings have to be measured in buildings, not assumed from its features.
Credit
a failure; US EPA; evaluators at Energy Market Innovations with Consumers Energy
2014
Rooftop units must report faults
Rooftop units are the boxed heating and cooling units on the roofs of shops and offices. Many have an economizer, a set of dampers that lets in cool outdoor air so the compressor can rest. Research by the New Buildings Institute found faults in 64% of the economizers it examined. From 1 January 2014, California's building energy code, Title 24, required automated fault detection and diagnostics (FDD) on new commercial units with an economizer and more than 4.5 tons of cooling capacity, the unit the US trade uses to size air conditioners. The controls must detect and report five faults: a failed air temperature sensor, not economizing when it should, economizing when it should not, a damper that does not move, and too much outdoor air.
Credit
California Energy Commission; Western Cooling Efficiency Center at UC Davis; the Western HVAC Performance Alliance
2017
Savings counted in real homes
ENERGY STAR returned to thermostats with a specification for connected (internet-linked) thermostats that took effect on 1 January 2017. EPA certifies a thermostat together with its maker's online service, using a year of data from hundreds of customers' homes run through EPA's own software, which it publishes as open source on GitHub. To qualify, the lower 95% confidence limit of the weighted national average must show at least 8% less heating run time and 10% less cooling run time. Utility Dive reported in March 2017 that Nest's learning thermostat was the first thermostat to earn the label since 2009.
Credit
US EPA ENERGY STAR
2025
Same unit, new firmware
The Northwest Energy Efficiency Alliance (NEEA), a group funded by utilities in the US Northwest, reported lab tests of two heat pumps of the same make, model and size from consecutive model years. Run against a simulated house load, the newer one worked at a coefficient of performance (COP, heat delivered per unit of electricity) of 3.37 against 2.03 for the older one, 66% higher, while its rated heating efficiency (HSPF) was only 8% higher, 13 against 12. NEEA said "the main difference was a new control algorithm".
LessonIn a variable-speed machine the control software can change real efficiency far more than the rating shows.
Credit
NEEA, with TRC and Bruce Harley Energy
Watch for
AI tunes warehouse HVAC
Trane Technologies, a maker of heating and cooling equipment, and Amazon Web Services said BrainBox AI, Trane's software that adjusts HVAC settings on its own, cut energy use by nearly 15% at three Amazon Grocery fulfillment centers in North America, more than double the project's target. They gave no baseline or method of measurement. They plan to extend it to more than 30 sites and to pilot it in grocery stores in 2026.
Achieved whenan outside party, such as a utility evaluator, publishes weather-adjusted savings measured at the electricity meter for those sites.
Claimed by
Trane Technologies and Amazon Web Services, December 2025
Open
01
An open thermostat test bench
A low-cost bench that tests any off-the-shelf home thermostat the same way and publishes the results by brand. A computer model of a house and heat pump decides the room temperature. A small heated and cooled box makes the thermostat's sensor feel that temperature. Inputs on the bench record which control wires the thermostat switches on: compressor stages, fan and backup electric heat. Scripted scenarios include warming up after a night setback, a cold morning when backup heat may come on, and a two-hour utility demand-response event.
The briefCredit · unclaimed02
Bench-test economizer fault alarms
An independent bench test of the economizer controllers with built-in fault detection that rooftop unit makers fit to meet rules such as California's. Mount each controller with its damper actuator and a small damper on a bench, feed its sensor inputs from adjustable signal sources, then create each fault one at a time: a failed or drifting sensor, a stuck damper, a damper that does not modulate. Record which faults it reports, how quickly, and any false alarms.
The briefCredit · unclaimed
Open 01 · the brief
An open thermostat test bench
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A low-cost bench that tests any off-the-shelf home thermostat the same way and publishes the results by brand. A computer model of a house and heat pump decides the room temperature. A small heated and cooled box makes the thermostat's sensor feel that temperature. Inputs on the bench record which control wires the thermostat switches on: compressor stages, fan and backup electric heat. Scripted scenarios include warming up after a night setback, a cold morning when backup heat may come on, and a two-hour utility demand-response event.
Why it is still open
ENERGY STAR certifies thermostats from each maker's own customer data, so no public test compares brands under the same conditions. The nearest work, at the National Renewable Energy Laboratory (NREL), connects thermostats to a real 3-ton heat pump and a simulated house; its 2025 study compared five unnamed thermostats over one simulated August day, in cooling only. NEEA's lab tests put a heat pump's thermostat in a "Thermostat Environmental Emulator" box, but they rate the heat pump, not the thermostat. Variable-speed heat pumps are a separate case: NEEA found in 2021 that only each maker's own "communicating" thermostats could run them as designed, so the bench covers conventional thermostats that switch 24-volt wires. Searches of the web and GitHub in November 2026 found software thermostat emulators but no open hardware bench with brand-by-brand results.
Built with what exists
a plug-in 24-volt AC transformer of the low-voltage kind sold for doorbells and thermostats; a microcontroller board with isolated inputs to read each control wire; a thermoelectric (Peltier) module, fan and reference temperature sensor in a small insulated box; a house and heat pump model in EnergyPlus or a simple heat-balance model in Python; Python to run the scenarios and log results.
First step
Wire one thermostat to a 24-volt transformer and four logged inputs, then warm and cool its sensor box by hand and record when it calls for heat and cooling.
Done looks like
At least six thermostat models from at least four brands, each run three times through at least eight published scenarios; a public table of compressor starts per hour, minutes of backup electric heat and degrees of overshoot for each; the parts list and code released, and a second group that builds the bench reproduces one thermostat's results within 10%.
Open 02 · the brief
Bench-test economizer fault alarms
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An independent bench test of the economizer controllers with built-in fault detection that rooftop unit makers fit to meet rules such as California's. Mount each controller with its damper actuator and a small damper on a bench, feed its sensor inputs from adjustable signal sources, then create each fault one at a time: a failed or drifting sensor, a stuck damper, a damper that does not modulate. Record which faults it reports, how quickly, and any false alarms.
Why it is still open
The Western Cooling Efficiency Center wrote in 2013 that California's rule would need lab testing to certify fault detection tools, and that a committee had begun ASHRAE Standard 207 to define that test. Lawrence Berkeley National Laboratory researchers wrote in 2020 that "there is a lack of standard methodology and datasets for evaluating the accuracy of FDD technologies"; their own test set covers air handlers, not rooftop unit economizers. In a 2022 ACEEE paper, Peter Jacobs and Robert Mowris had four packaged units with economizers tested at Intertek, a commercial lab, and found that the economizers let in 18% outdoor air on average with the dampers closed and 41% at their minimum setting, leaving the units 38 to 62% less efficient than their AHRI ratings; they measured airflow and efficiency, not fault alarms, and wrote that Standard 207 tests fault detection but sets no performance standard for economizers on packaged units. Searches in November 2026 found no published results of economizer controllers tested to Standard 207.
Built with what exists
economizer controllers with built-in fault detection, bought from HVAC parts suppliers, with their sensors and damper actuators; a plug-in 24-volt transformer; resistor boxes or a programmable signal source to stand in for temperature sensors; a microcontroller to log alarms; ASHRAE Standard 207, bought from ASHRAE, for the fault list and pass rules.
First step
Buy one controller with its sensors and actuator, power it on a bench, and log its alarm output while disconnecting the outdoor air sensor.
Done looks like
At least four controllers from different makers, each put through every fault at least five times; a published table of faults detected, minutes to alarm, and false alarms during a 72-hour run with no faults, with the bench design released so others can repeat it.